luissattelmayer/policy-politics

tmpyxthtyno

This model is a fine-tuned version of xlm-roberta-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2418
• F1: 0.9725
• Precision: 0.9690
• Recall: 0.9760

lang	f1	precision	recall	prevalence	n
cy	1.000000	1.000000	1.000000	0.812500	16
hu	1.000000	1.000000	1.000000	0.529412	17
sv	0.987552	0.975410	1.000000	0.868613	137
no	0.980392	1.000000	0.961538	0.666667	39
de	0.977853	0.972881	0.982877	0.630670	463
da	0.974619	0.989691	0.960000	0.800000	125
en	0.974138	0.974138	0.974138	0.758170	153
it	0.971429	0.980769	0.962264	0.670886	79
fr	0.971429	0.944444	1.000000	0.566667	30
es	0.967742	0.937500	1.000000	0.714286	21
nl	0.966507	0.961905	0.971154	0.675325	154
fi	0.960000	0.923077	1.000000	0.750000	16
pl	0.936170	0.956522	0.916667	0.592593	81
cs	0.933333	0.875000	1.000000	0.411765	17
pt	0.833333	0.714286	1.000000	0.454545	22

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 8
• eval_batch_size: 16
• seed: 1234
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	F1	Precision	Recall
0.2087	0.9989	437	0.1219	0.9751	0.9806	0.9696
0.1227	2.0	875	0.1591	0.9724	0.9868	0.9584
0.0943	2.9989	1312	0.1554	0.9744	0.9759	0.9728
0.0465	4.0	1750	0.2055	0.9732	0.9868	0.96
0.0339	4.9943	2185	0.1788	0.9759	0.9791	0.9728

Framework versions

• Transformers 4.45.1
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.20.3